transform. Change some variable names so it is clear what is source and
what is dest of the cast. Also, add an assert to ensure that the integer
to integer case is asserting if the bitwidths are different. This prevents
illegal casts from being formed and catches bitwidth bugs sooner.
llvm-svn: 33337
not to overflow 64-bits and end up with a 0 mask. This caused i64 values to
always be stored as 0 with lots of consequential damage to nightly test.
llvm-svn: 33335
mostly be used by back ends that wish to distinguish between integer types
that fit evenly within a natural integer type and those that don't.
llvm-svn: 33328
1. Fix logic for executeGEP. Only 32-bit and 64-bit integer types are
acceptable as indices.
2. Ensure that all integer cast operations truncate their result to the
integer size of the operand.
llvm-svn: 33318
Add the llvm.stacksave and llvm.stackrestore intrinsics that correspond
to the GCC builtins __builtin_stack_save and __builtin_stack_restore. This
allows the CBE to execute these builtins, and fixes:
test/CodeGen/CBackend/2007-01-17-StackSaveNRestore.ll
External/SPEC/CINT2000/176.gcc (CBE)
Patch by Gordon Henriksen.
llvm-svn: 33313
Ensure the CBE generates calls to GCC builtins for stack_save and
stack_restore when it encounters the corresponding intrinsics.
Patch by Gordon Henriksen.
llvm-svn: 33312
Support overriding the prefix and DESTDIR variables from the make command
line. This makes the makefile system a little more GNU compliant and the
feature makes it easier for distribution builders to re-target the install
location of LLVM.
llvm-svn: 33305
non-statics.
* Introduce new option to output zero-initialized data to .bss section.
This can reduce size of binaries. Enable it by default for ELF &
Cygwin/Mingw targets. Probably, Darwin should be also added.
llvm-svn: 33299
Make the assembler generate a nice error message if a bad cast instruction
is attempted instead of asserting out. This is made possible by the
recently exposed method CastInst::castIsValid() which checks the validity
of any cast instruction.
llvm-svn: 33283
careful when folding "c ? load p : load q" that C doesn't reach either load.
If so, folding this into load (c ? p : q) will induce a cycle in the graph.
llvm-svn: 33251
LLVM would miscompile ASM dialects when compiling for PPC. Added dialects for
the X86 and PPC backends. It defaults to "0", the first variant of a compound
inline asm expression.
llvm-svn: 33246
because TargetData::getTypeSize() returns the same for i1 and i8. This fix
is not right for the full generality of bitwise types, but it fixes the
regression.
llvm-svn: 33237
the basic block and is stable across runs in gdb or valgrind.
Make Node::update handle edges which dominate and are tighter than
existing edges.
Replace makeEqual's "squeeze theorem" code. Fixes miscompilation.
Gate the calls to defToOps and opsToDef. Before this, we were getting IG
edges about values which weren't even defined in the dominated area. This
reduces the size of the IG by about half.
llvm-svn: 33236
instcombine to clean up the code, which makes more code suitable for SRoA.
This helps C++ code in particular, e.g. speeding up tramp3d by 31%.
llvm-svn: 33235
Increment the counter after the second use, not after the first use. This
fixes PR1113.
Also, rename some classes for simplicity and to more naturally be
reminscient of LLVM 1.9. This in preparation for additional classes that
will provide a scaled down model of the LLVM 1.9 IR.
llvm-svn: 33226
rename Type::getIntegralTypeMask to Type::getIntegerTypeMask.
This makes naming much more consistent. For example, there are now no longer any
instances of IntegerType that are not considered isInteger! :)
llvm-svn: 33225
which includes i1. Change the semantics of the various instructions to
support i1 where they didn't before: for example, it is now legal [though
not terribly useful, as it is the same as xor] to add two i1 values.
llvm-svn: 33217
like a lot, its really only two changes:
1. Move stuff that should be private to .y out of the .h file.
2. Make all semantic values pointers.
This cleans up the silly destroy methods and changes them to delete. It
also moves the TypeInfo and ValueInfo classes into the .y where we can
modify them more readily because they have no clients any more.
This shouldn't result in any functional changes in llvm-upgrade.
llvm-svn: 33211
Remove now-extraneous checks for ptr->ptr bitcasts.
Fix PR1109 and Analysis/BasicAA/2007-01-13-BasePointerBadNoAlias.ll. We
need to consider arbitrary sized objects when checking for nested GEP offsets.
llvm-svn: 33195
Bye, Bye Booly. Remove the use of the bool type from non-upgraded test
cases and from grep expressions. The parser doesn't accept it and the
asm writer doesn't produce it any more.
llvm-svn: 33183
that properties were being applied where they didn't belong. Fixes crash
in new MiBench testcase.
Also mark debugging code as such in #ifdef.
llvm-svn: 33177
Implement upgrading of the varargs intrinsics. These must now have i8*
arguments. This patch bitcasts arguments to i8* and fixes the prototypes.
llvm-svn: 33163
* PIC-aware internal structures in X86 Codegen have been refactored
* Visibility (default/weak) has been added
* Docs fixes (external weak linkage, visibility, formatting)
llvm-svn: 33136
(hasn't for ages) so move it up one level in the table of contents.
Type needs to be understood before Value so move it before Value. Make
the descriptions of types stand out a little more. Add references to the
doxygen for the Type class.
llvm-svn: 33122
1. int -> i32
2. Describe the IntegerType class.
3. Correct the description of Type and its primitive type subclasses.
4. Document OpaqueType and PackedType a little better.
llvm-svn: 33120
Implement the arbitrary bit-width integer feature. The feature allows
integers of any bitwidth (up to 64) to be defined instead of just 1, 8,
16, 32, and 64 bit integers.
This change does several things:
1. Introduces a new Derived Type, IntegerType, to represent the number of
bits in an integer. The Type classes SubclassData field is used to
store the number of bits. This allows 2^23 bits in an integer type.
2. Removes the five integer Type::TypeID values for the 1, 8, 16, 32 and
64-bit integers. These are replaced with just IntegerType which is not
a primitive any more.
3. Adjust the rest of LLVM to account for this change.
Note that while this incremental change lays the foundation for arbitrary
bit-width integers, LLVM has not yet been converted to actually deal with
them in any significant way. Most optimization passes, for example, will
still only deal with the byte-width integer types. Future increments
will rectify this situation.
llvm-svn: 33113
recommended that getBoolValue be replaced with getZExtValue and that
get(bool) be replaced by get(const Type*, uint64_t). This implements
those changes.
llvm-svn: 33110
in getLocalSlot fires on many, many values. It broke nearly all of
the dejagnu tests. Simple changes to the assertion did not fix the
problem.
llvm-svn: 33054
Invert the "isSigned" logic in calls to printType and printPrimitiveType.
We want variables to be declared unsigned by default so that signless
operators like + and - perform the unsigned operation that LLVM expects
by default. Parameters with the sext attribute will be declared signed and
signed instructions will case operand values to signed regardless of the
type of the variable. This passes all tests and fixes PR1099.
llvm-svn: 33039
Partial fix for this PR. Default function parameters to signed integer, just
like everything else in CBE. The bug was caused by incorrectly introducing
parameter attributes feature by choosing "signed" parameter if the
SExtAttribute was specified. Howeer, if no attribute is specified, this
causes it to become unsigned which is incorrect. Reversing the logic so
that signedness is detected by "not ZExtAttribute" set fixes the issue.
This fixes 197.parser but there is more to do. Any comparison and possibly
other operators involving arguments may need to correctly cast the parameter
before its use, depending on the sign of the operator.
llvm-svn: 33034
- Fixed bugs in sethi-ullman number computation and priority queue comparison
functions.
- Separate code that handles priority computation special cases from SU number computation.
llvm-svn: 33025
Eventually, Top level pass managers will use this to keep track of
active pass managers. Eass pass will also learn how to find appropriate
manager from these managers stack.
llvm-svn: 33018
This patch converts getPrimitiveSize to getPrimitiveSizeInBits where it is
appropriate to do so (comparison of integer primitive types).
llvm-svn: 33012
Enable complex addressing modes on 64-bit platforms involving two induction
variables by keeping a size and scale in 64-bits not 32.
Patch by Dan Gohman.
llvm-svn: 33011
Clean up the definitions of the helper functions per Chris' review
suggestions so they are easier to read.
For PR1091:
Print minimum signed integer values as unsigned so that we get no warnings
from the C compiler about constant ranges and value comparisons.
llvm-svn: 33010
Implemented some llvm_fcmp_{pred} functions at the start of the function bodies
and use them for fcmp instructions and constant expressions. These help
implement the ordered and unordered comparisons necessary for correct exectuion
of these comparisons.
llvm-svn: 33007
Take an incremental step towards type plane elimination. This change
separates types from values in the symbol tables by finally making use
of the TypeSymbolTable class. This yields more natural interfaces for
dealing with types and unclutters the SymbolTable class.
llvm-svn: 32956
Solve several related problems by making variable names more unique and
dealing with recursive phi nodes. Unfortunately, this doesn't solve the
main issue reported in the PR, but its a step in that direction.
llvm-svn: 32953
Emit the correct syntax for icmp and fcmp constant expressions. The
predicate goes outside the parentheses not inside. This was a change made
in the assembler without the corresponding change made in llvm-upgrade.
llvm-svn: 32942
- Add isSelectExpensive() etc. It's used to tell codegen that select is expensive for a given target, avoid using it if possible. Currently it's only
used to expand FCOPYSIGN.
llvm-svn: 32939
Disallow merging of dupliate global variables. It is now illegal to declare
or define two global variables of the same name and same type. llvm-gcc3 is
dead in 2.0 and llvm-gcc4 doesn't have that problem nor need the hack.
llvm-svn: 32933
or'ing in the sign bit of operand 1.
- Tweaking: rather than left shift the sign bit, fp_extend operand 1 first
before taking its sign bit if its type is smaller than that of operand 0.
llvm-svn: 32932
conflict after upgrade resulting from collapsed type planes. The test now
checks to make sure llvm-upgrade produces appropriate warning messages.
llvm-svn: 32913
Split this test case into three cases. globalredefinition.ll just tests
that a global can be forward referenced. globalredefinition2.ll tests
that llvm-upgrade will warn about renaming duplicate global variables.
globalredefinition3.ll tests that llvm-as will generate an error on
duplicate global variables. This last test is currently XFAIL because
the restriction is not implemented yet.
llvm-svn: 32910
1. Completely revise the type system so that types are handled as const
objects and not created multiple times, cloned, or otherwise copied.
This gets around memory issues, saves memory, and also emulates LLVM's
no-two-types-of-the-same-shape-created semantics.
2. Adjust the handling of global names. Basically, we cannot rename them
for a variety of reasons: linking, forward references, etc.
3. Detect global names that have name conflicts as the result of collapsed
type planes or redefinitions that llvm-as no longer accepts. These will
produce warnings on stderr and one of the globals will be renamed.
4. Rename ParserInternals.h as UpgradeInternals.h so it doesn't conflict
in the debugger with ParserInternals.h from lib/AsmParser.
5. Move the guts of the TypeInfo class into the grammar so we aren't
implementing large functions in a header file. This also helps with
debugging a bit.
llvm-svn: 32906
1. The @ sign is no longer necessary.
2. We now support "function attributes" as parameter attribute 0.
3. Instead of locating the return type attributes after the type of a
function result, they are now located after the function header's
closing paranthesis and before any alignment or section options.
4. The way has been prepared for a new "noreturn" function attribute but
there is no support for recognizing it in the lexer nor doing anything
with it if it does get set.
5. The FunctionType::getParamAttrsText method now has support for
returning multiple attributes. This required a change in its interface.
I'm unhappy that this change leads to 6 new shift/reduce conflicts, but
in each case bison's decision to choose the shift is correct so there
shouldn't be any damage from these conflicts.
llvm-svn: 32904
by correctly handling up references and enumerated types. Previously there
was some confusion of these two. Thanks to Chris Lattner for demistifying
llvm up references.
llvm-svn: 32846
- New target type "mingw" was introduced
- Same things for both mingw & cygwin are marked as "cygming" (as in
gcc)
- .lcomm is supported here, so allow LLVM to use it
- Correctly use underscored versions of setjmp & _longjmp for both mingw
& cygwin
llvm-svn: 32833
things: (1) preventing PR1071 and (2) working around missing parameter
attributes for bool type. (2) will be fixed shortly. When PR1071 is fixed,
this patch should be undone.
llvm-svn: 32831
1. Switch expression and cases are compared signed and are sign extended.
2. For function results needing extended, do SIGN_EXTEND if the SExtAttribute
is set and ZERO_EXTEND if the ZExtAttribute is set, otherwise just let
the Legalizer do ANY_EXTEND.
This fixes the recent regression in kimwitu++ and probably the llvm-gcc
bootstrap issue we had today.
llvm-svn: 32830
it used an undefined type name as a parameter argument. This bug in the
assembler has been fixed and it is now necessary to define the type.
llvm-svn: 32826
to MRegisterInfo. These allow the target to issue instructions to spill and
restore callee saved registers in case where individual stores / loads aren't
the correct / profitable choice.
llvm-svn: 32820
Tweak the RUN: lines of these tests to accommodate the renaming of
variables done by llvm-upgrade. The renaming occurs as a result of avoiding
name collisons for collapsed type planes. Conflicting names have a .u
(unsigned) or .s (signed) suffix added. This patch updates the grep
expression to accommodate the new names.
llvm-svn: 32815
Add a test case to ensure that llvm-upgrade retains correct semantics
for the previous definition of "cast to bool" which compared against null.
llvm-svn: 32812
Revise the upgrade parser to keep track of types more faithfully and use
this information to resolve name conflicts resulting from collapsed type
planes. The type planes have collapsed because the integer types are now
signless so that uint and int became i32. Where two planes existed for uint
and int, only i32 exists. Any variable names depending on the type planes
to pmake the identifier unique would cause a conflict. This patch resolves
that conflict for many but not all cases.
Situations involving the integer types and pointers to them are handled
by this patch. However, there are corner cases that are not handled
well, such as:
%t1 = type { uint, int }
%t2 = type { int, uint }
void %myfunc(%t1* one, %t2* two) {
%var = load %t1* one
%var = load %t2* two
}
In the scenario above, %t1 and %t2 are really the same type: { i32, i32 }
Consequently attempting to name %var twice will yield a redefinition error
when assembled.
While this patch is sufficien to allow the llvm/test suite to pass, More
work needs to be to complete the handling of these corner cases.
llvm-svn: 32810
Update for signless integer types:
1. Replace [us]byte with i8
2. Replace [u]short with i16
3. Replace [u]int with i32
4. Replace [u]long with i64
5. Document the "define" keyword and use it in all examples.
6. Document parameter attributes and how they affect function types.
llvm-svn: 32791
Update the test suite to accommodate the change from signed integer types
to signless integer types. The changes were of only a few kinds:
1. Make sure llvm-upgrade is run on the source which does the bulk of the
changes automatically.
2. Change things like "grep 'int'" to "grep 'i32'"
3. In several tests bitcasting caused the same name to be reused in the
same type plane. These had to be manually fixed. The fix was (generally)
to leave the bitcast and provide the instruction with a new name. This
should not affect the semantics of the test. In a few cases, the
bitcasts were known to be superfluous and irrelevant to the test case
so they were removed.
4. One test case uses a bytecode file which needed to be updated to the
latest bytecode format.
llvm-svn: 32789
Three changes:
1. Convert signed integer types to signless versions.
2. Implement the @sext and @zext parameter attributes. Previously the
type of an function parameter was used to determine whether it should
be sign extended or zero extended before the call. This information is
now communicated via the function type's parameter attributes.
3. The interface to LowerCallTo had to be changed in order to accommodate
the parameter attribute information. Although it would have been
convenient to pass in the FunctionType itself, there isn't always one
present in the caller. Consequently, a signedness indication for the
result type and for each parameter was provided for in the interface
to this method. All implementations were changed to make the adjustment
necessary.
llvm-svn: 32788
This patch replaces signed integer types with signless ones:
1. [US]Byte -> Int8
2. [U]Short -> Int16
3. [U]Int -> Int32
4. [U]Long -> Int64.
5. Removal of isSigned, isUnsigned, getSignedVersion, getUnsignedVersion
and other methods related to signedness. In a few places this warranted
identifying the signedness information from other sources.
llvm-svn: 32785
Don't attempt to parse both the old and new grammars. It is near impossible
to get it right. Remove support for the new define keyword and don't
attempt to insert parameter attributes because there isn't enough
contextual information for it.
llvm-svn: 32784
Update for signless integer types and parameter attribute implementation.
Of significant note:
1. This changes the bytecode format yet again.
2. There are 1/2 as many integer type planes (this is a good thing)
3. GEP indices now use only 1 bit to identify their type which means
more GEP instructions won't be relegated to format 0 (size win)
4. Parameter attributes are implemented but currently being stored
verbosely for each function type. Some other day this needs to be
optimized for size.
llvm-svn: 32783
Major reorganization. This patch introduces the signedness changes for
the new integer types (i8, i16, i32, i64) which replace the old signed
versions (ubyte, sbyte, ushort, short, etc). This patch also implements
the function type parameter attributes feature. Together these conspired
to introduce new reduce/reduce errors into the grammar. Consequently, it
was necessary to introduce a new keyword into the grammar in order to
disambiguate. Without this, yacc would make incorrect shift/reduce and
reduce/reduce decisions and fail to parse the intended assembly.
Changes in assembly:
1. The "implementation" keyword is superfluous but still supported. You
can use it as a sentry which will ensure there are no remaining up
reference types. However, this is optional as those checks are also
performed elsewhere.
2. Parameter attributes are now implemented using an at sign to
indicate the attribute. The attributes are placed after the type
in a function declaration or after the argument value in a function
call. For example:
i8 @sext %myfunc(i16 @zext)
call i8 @sext %myfunc(i16 @zext %someVal)
The facility is available for supporting additional attributes and
they can be combined using the @(attr1,attr2,attr3) syntax. Right
now the only two supported are @sext and @zext
3. Functions must now be defined with the "define" keyword which is
analagous to the "declare" keyword for function declarations. The
introduction of this keyword disambiguates situations where a
named result type is confused with a new type or gvar definition.
For example:
%MyType = type i16
%MyType %func(%MyType) { ... }
With the introduction of optional parameter attributes between
the function name and the function result type, yacc will pick
the wrong rule to reduce unless it is disambiguated with "define"
before the function definition, as in:
define %MyType @zext %func(%MyType %someArg) { ... }
llvm-svn: 32781
Add a new feature to FunctionType, Parameter Attributes. This allows tags
such as "sext" and "zext" to be associated with a faunction's arguments
or return type. This allows signedness information to be carried forward
from the frontend to the backend for arguments and result types.
llvm-svn: 32776
depend on the compiler. This works around problems in the Stacker runtime
when the CFE changes in such a way that the assembly file needs to be
updated.
llvm-svn: 32773
Fix several bugs and update for new assembly syntax. Changes made include:
1. Fixing rules for icmp/fcmp instructions to not require a closing paren
at the end. This was a cut-and-paste error from a previous commit.
2. Changing things like Out << " " to Out << ' '
3. Adding the "define" keyword for function definitions
4. Adding support for packed structures
llvm-svn: 32771
Remove all grammar conflicts from assembly parsing. This change involves:
1. Making the "type" keyword not a primitive type (removes several
reduce/reduce conflicts)
2. Being more specific about which linkage types are allowed for functions
and global variables. In particular "appending" can no longer be
specified for a function. A differentiation was made between the various
internal and external linkage types.
3. Introduced the "define" keyword which is now required when defining a
function. This disambiguates several cases where a named function return
type could get confused with the definition of a new type. Using the
keyword eliminates all shift/reduce conflicts and the remaining
reduce/reduce conflicts.
These changes are necessary to implement the function parameter attributes
that will be introduced soon. Adding the function parameter attributes in
the presence of the shift/reduce and reduce/reduce conflicts led to severe
ambiguities that caused the parser to report syntax errors that needed to
be resolved. This patch resolves them.
llvm-svn: 32770
1. Make the arguments const like the other ConstantFold* functions.
2. Clean up evaluateFCmpRelation so it makes sense for floating point.
3. Implement the use of evaluateFCmpRelation to fold floating point CEs
4. Shorten a variable name so more things fit on one line.
5. Fix various comments.
llvm-svn: 32759
Fix this by ensuring that a bitcast is inserted to do sign switching. This
is only temporarily needed as the merging of signed and unsigned is next
on the SignlessTypes plate.
llvm-svn: 32757
This patch removes the SetCC instructions and replaces them with the ICmp
and FCmp instructions. The SetCondInst instruction has been removed and
been replaced with ICmpInst and FCmpInst.
llvm-svn: 32751
pass managers. Otherwise, stale available analysis info, from the managers not
yet run, may cause pass manager to take wrong turn.
This fixes CBE test failures reported by nightly tester.
llvm-svn: 32726
John Criswell